An Overview of Dexterous Manipulation

Allison M. Okamura, Niels Smaby and Mark R. Cutkosky

Dexterous Manipulation Laboratory
Stanford University

touch@cdr.stanford.edu

man grasp choices for particular objects and tasks,

other researchers have argued that robot hands, and
the circumstances in which they work, are fundamentally dierent from the human condition. A modelbased approach, based on the kinematics and dynamics of manipulating an object with the ngertips, has
therefore dominated the eld. The results of this approach are now adequate for manipulations of objects
in controlled environments. As discussed in Section 6,
the main limitation appears to be a lack of adequate
tactile sensing for robust manipulation control.
Examples of autonomous, robotic dexterous manipulation are still conned to the research laboratory.
However, the model-based approach has already provided considerable insight into the nature of dexterous
manipulation, both in robots and in humans. Some
of these results are now being applied to reconstructive surgery, in which hand surgeons perform tendon
transfer surgeries on patients with quadriplegia and/or
nerve palsies to improve their grasping ability [39].
Future applications of robotic dexterous manipulation may include tasks where ne manipulation is required, yet it is infeasible or dangerous for a human to
perform the task. Examples include underwater salvage and recovery, remote planetary exploration, and
retrieval of objects from hazardous environments.

Abstract

This paper for the ICRA 2000 Symposium on Dexterous Manipulation presents an overview of research
in dexterous manipulation. We rst dene robotic
dexterous manipulation in comparison to traditional
robotics and human manipulation. Next, kinematics,
contact types and forces are used to formulate the dexterous manipulation problem. Dexterous motion planning is described, which includes grasp planning and
quality measures. We look at various mid- and lowlevel control frameworks, and then compare manipulation versus exploration. Finally, we list what we see as
the current limiting factors in dexterous manipulation,
and review the state of the art and future of the eld.

1 Introduction

Dexterous Manipulation is an area of robotics in

which multiple manipulators, or ngers, cooperate to
grasp and manipulate objects. Dexterous manipulation diers from traditional robotics primarily in that
the manipulation is object-centered. That is, the problem is formulated in terms of the object to be manipulated, how it should behave, and what forces should
be exerted upon it. Dexterous manipulation, requiring precise control of forces and motions, cannot be
accomplished with a conventional robotic gripper; ngers or specialized robotic hands must be used.
Although humans are not the only creatures capable of manipulation, it is quintessentially a human activity. The fraction of the human motor cortex devoted to manipulation and the number and sensitivity
of mechanoreceptors in our palms and ngertips are
indications of the importance of manipulation in humans. Within a year of birth, a human infant is clearly
more dexterous than today's robots, though far short
of adult skill in grasping and handling objects.
It should come as no surprise then that the majority of robot hands designed for dexterous manipulation are anthropomorphic in design. Research has also
been done to classify human grasping and manipulation with an eye to providing a knowledge-based approach to grasp choice for robots [5, 6, 23, 15]. While
this approach has had some success in emulating hu-

Formulation of the Dexterous

Manipulation Problem

The rst step in moving an object from one conguration to another using robotic ngers is to formulate the dexterous manipulation (DM) problem (Figure 1). This problem sets the framework for determining the required actuator forces/torques to produce
the desired motions of the object. In keeping with an
object-centered approach, we work \backwards" from
the object to the manipulators. The development of
the kinematic portion of the DM problem, done here
from force/torque relationships, can also be accomplished from linear and angular velocity relationships.
This development requires knowing the geometric
relationships of the dexterous manipulator-object
system (i.e., contact locations), object geometry,
ngertip and link geometry, and the kinematics of the
1

Joint

object
coordinate
frame

1
2

Contact
T

Force/Torque
Space

space

Jh

Velocity
Space

space

Jh

fc
space

Object

f
m

B
A
contact
coordinate
frame

manipulator. In addition, it is assumed that contact

is maintained throughout the manipulation.

2.1.1

Jacobian Relationships

The rst step in developing the kinematics of the DM

problem is to calculate the required ngertip forces
from a desired force/torque wrench on the object. The
basis for this calculation is the grasp Jacobian relationship.

= Ji T ftip
(4)
In the DM problem, these individual Jacobians are
brought together to form the hand Jacobian.

= Gftip
(1)
The grasp Jacobian (or grasp map), G, can be obtained by resolving each ngertip force to a common
coordinate frame embedded in the object. For each
ngertip i, this force resolution results in the mapping
matrix Gi .

i

f obj

3
2
1
J1T
6 2 7
6 0
6
7
6
4 ::: 5 = 4 :::
m
0

= Gi ftip
(2)
In Equation 2, the force vectors f are generalized
vectors: they may include both forces and torques.
The individual mapping matrices Gi are concatenated
to form the grasp map G, and the ngertip force vectors are also grouped into one vector.
i

2
fobj

G1

G2

:::

Gm

6
6
4

ftip1
ftip2

:::
ftipm

:::
:::
:::

J2T
:::
:::

32

54

6
::: 7
76

T
Jm

ftip1
ftip2

3
7
7

::: 5
ftipm

(5)

A nice conceptual picture of the roles of the grasp

Jacobian and the hand Jacobian is shown in Figure 2
[25]. Given a set of contact forces, the individual joint
torques can be obtained by multiplying by the transpose of the hand Jacobian, Jh , and the forces on the
object can be obtained by multiplying by the grasp
Jacobian, G.

3
7
7
5

minimizing the dimension of the problem. For a detailed treatment of this topic see [19, 28, 30].
The grasp Jacobian developed above allows us to
calculate the required contact forces from the desired
force on the object. In order to produce these forces at
the ngertips, we now develop a hand Jacobian, which
will allow us to calculate the joint torques from the
contact forces [19]. The hand Jacobian, Jh , is based
on the standard Jacobian, which relates end eector
forces to individual joint torques for a robotic manipulator (in this case one for each nger).

Kinematics

fobji

Figure 2: The role of the hand and grasp Jacobians.

For the ngers,  is the vector of joint torques and _ is
the vector of joint velocities. For the contacts, fc is the
vector of contact forces and x_ c is the vector of contact
point velocities. For the object, the resultant force
vector and the vector of object velocities are shown.

Figure 1: A typical dexterous manipulation problem:

Moving an object from conguration A to conguration B .

2.1

xc
space

(3)

fobj

Note that Equation 3 is a simplied treatment of

the problem. Typically the ngertip forces are represented in a coordinate frame at the contact point on
the surface of the object. Then, knowing each contact
type (see the sections below and [22]), the number of
allowable force directions at each contact is reduced,

=
=

Jh T fc
Gfc

(6)
(7)

Alternatively, from the kinematic point of view, the

contact point velocities can be obtained from the nger
joint velocities by multiplying by the hand Jacobian or
by multiplying the object velocity by the grasp Jacobian:
2

Relative velocities
of contact points on
each of the two objects

Jh _ = x_ c

= GT vobj
(8)
The kinematic and force relationships described by
the above equations and Figure 2 are not necessarily
one to one. The system may be over-constrained (i.e.
the ngers may not be able to accommodate or resist all object motions or forces) or the system may
be under-constrained (i.e. there are multiple choices
for nger joint velocities or torques). Typically, an
under-constrained system is desired for dexterous manipulation tasks. The detection of these conditions can
be accomplished by treating the combination of ngers
and object as a parallel-chain mechanism and evaluating the manipulability of the object with respect to the
palm [19]. A summary of kinematic measures useful
in dexterous manipulation is provided in [6].
2.1.2

vx
vy
x
y
z

Velocities of contact
frames on objects
(1 and 2) and spin

Contact
Equations

}u
}u

Figure 3: Contact variables (adapted from [29]).

Rolling and Sliding

The previous development of the kinematics is for

point contacts on the object, which do not move during the manipulation. However, the geometry of typical robotic ngers causes the contact points to travel
on the surface of the object as the conguration of
the ngers change during the manipulation. This is
typically a non-holonomic constraint: the rolling of a
ngertip on an object requires that the velocities of the
contact points on each of the two objects must remain
the same. Including rolling and sliding constraints in
the DM problem kinematics involves the application of
dierential geometry and a parameterization of both
the ngertip and object surfaces. This analysis takes
as inputs the relative velocities (linear and angular) of
the contact points on the objects (in this case on the
nger and the object), and outputs the parameterized
contact point velocities on the surfaces of the objects
(Figure 3). Figure 4 shows a typical example of the
progression of the nger and object contact coordinate
frames for the rolling of a nger on the surface of an
object. For a detailed treatment of the dierential geometry involved in rolling see [29].
We will now look at the contact constraints and
reduce a general rolling and sliding problem to pure
rolling in the contact plane. When the only constraint
is to maintain contact, the relative velocity of the contact points can have no component in the surface normal direction (vz = 0); this is assumed for most DM
problems. Adding the rolling constraint means that
there can be no relative linear velocity between the
two contact points, therefore vx = 0 and vy = 0. Finally, if we specify soft nger contact (we do not allow
spin of the ngertip about the contact normal) in the
plane, there are two more constraints on angular velocity (!x = 0 and !z = 0). A summary of the velocity
constraints for pure rolling in the plane is shown in
Table 1.
However, dexterous manipulation is not conned to
pure rolling. Slip often occurs, and is useful when ex-

Figure 4: The contact frames for rolling. u1 and u2

are the contact frames for objects 1 and 2, respectively.
After rolling occurs, the new contact frames will be u01
and u02 .
ploring an unknown object or when changing the pose
of a grasp to maintain control of the object [34]. One
diculty with slip is that ngertip contact sensors are
often necessary to determine contact location, whereas
with pure rolling one can determine the current contact location from relative motions and starting contact locations. For a more detailed treatment of sliding
manipulation see [18, 35].
2.2

Contact Types and Force Closure

Early in the study of dexterous manipulation it

was recognized that the kinematics and dynamics are
strongly in uenced, even dominated, by the contact
conditions at the ngertips [7]. At a basic level, there
are three representative contact types: point contact
without friction, point contact with friction, and soft
nger contact (Figure 5). The point contact without
friction can only resist a unidirectional force normal
to the surface. Adding friction will allow it to resist
Maintain contact
No sliding
No spin
Planar rolling

vx

vz = 0
= 0, vy = 0
!z = 0
!x = 0

Table 1: Contact velocity constraints for pure rolling

in the plane
3

Frictionless
point contact

closure grasp and a \manipulable" grasp. Force closure requires only that the ngers can resist an externally applied force, i.e. the opposing resistive force
can be passive or structural. A manipulable grasp requires that the manipulator can actively accommodate
all object motion directions while maintaining contact.

Fn
Ftx

Point contact
with friction

2.3

Fn

If frictional forces are relied upon to achieve a stable grasp, it is important to provide sucient contact
normal forces. To increase these normal forces, an internal force is supplied. This is a vector of contact
force magnitudes that impart no resultant force to the
object and thus lie in the null space of the grasp map,
G. If a vector of contact force magnitudes is decomposed into those producing external forces and those
producing internal forces, each separate force magnitude vector must satisfy the unidirectional constraints
required by the contact types (i.e. ngers can only
push, not pull, on the object surface). In general,
there are many solutions of grasp forces that satisfy
grasp stability while keeping each contact force inside
its friction limits and supplying some internal force.
This leads to an optimization problem, as discussed
in the next section. One treatment of this problem is
the \virtual linkage" concept developed in [41], where
virtual links are imagined between each unique pair of
contact points. The internal forces are then the forces
that each virtual link would experience during the manipulation. For other, more detailed treatments of the
internal/external force concept see [42, 28].

Fty
Ftx

Soft-finger
contact

Mz

Fty

Fn

Figure 5: Common contact types and their associated

forces/moments.
No Slip

Fc

Slip

Friction
Cones

Fc

Figure 6: A representation of contact forces, with and

without slip.

tangential forces and the soft nger can, in addition,

resist a torque about the normal to the surface.
When analyzing grasps which involve frictional contacts, it is also important to consider the friction cone.
This is a geometric representation of the frictional
force limits due to the static coecient of friction. In
order for no slippage to occur, the contact force must
lie within the friction cone (Figure 6). For a soft nger, the friction cone can be replaced by a limit surface that includes torsional friction [12]. For a more
detailed treatment of contact types and force closure
see references [28, 22].
Choosing contact locations is an important part of
dexterous manipulation. Typically, it is important to
achieve force closure on an object when manipulating
it (there are also important non-force closure manipulations - pushes, tips, etc.). Force closure requires that
the grasp of the object can resist any possible direction of force/torque perturbation to the object. When
a grasp has force closure on an object it is referred to
as being a stable grasp.1
It is also important to distinguish between a force
1 There

Internal and External Forces

Dexterous Motion Planning

In dexterous motion planning, there are two main

objectives to consider: planning the motion of the object to achieve a desired conguration or accomplish a
task, and planning the grasp or motion of the ngers
required to impart this motion. This discussion overlaps with the grasp planning and optimization work
reviewed in the Contact and Grasp Symposium at this
conference [22].
3.1

Grasp Planning for Desired Object

Motion

There are two main questions that can be used to

address the problems of grasp and grasp gait planning.
These are: How can a sequence of local motions and
regrasps be found which will result in the nal desired
object motion? Is there a guarantee that a new stable grasp will always exist given a specic object geometry, nger workspace and grasp such that motion
can continue? These questions are answered using the
kinematics of the robot hand, a reachability analysis,
and consideration of regrasping and gaiting.

are also other denitions of a stable grasp [6].

High-Level Task planning, discrete event

systems, grasp choice
Mid-Level Phases, transitions, event detection
Low-Level Operational space dynamics, cooperative object impedance control, kinematics, forces

Given a desired object motion, one must rst determine whether the ngers can move the object without
regrasping. This can be accomplished by a reachability analysis [19]. Using the range of possible manipulations for a given hand and object, a workspace can
be constructed that is a function not only of the geometry of the hand and object, but also of the rolling or
sliding that may occur at the contacts. If the ngers
must disconnect with the object in order to move the
object to the desired position, then one can consider
regrasping. A sequence of nger motions and regrasps
are known as a grasp gait [24].
When moving or reorienting an object with a hand,
only a nite amount of local motion can be imparted
to the object before a new grasp must be found. This
can be due to the limited workspace of the ngers of
any hand (human or robot), or collisions among the
nger links, the environment and the object being
manipulated. One method to determine whether local
motions will suce to reorient the object is the grasp
map, a graphical representation of all stable grasps
[24]. In planning, it is also important to realize that
a new grasp cannot always be found if the object
is moved locally until a nger reaches a workspace
limit; often a grasp gait must occur before the limit is
reached.
3.2

Grasp Optimizations
Measures

and

Table 2: Levels of Control for Dexterous Manipulation

typically a non-convex (and non-linear) function over
the search space and thus standard convex optimization techniques can not be used. There have been attempts at contact location synthesis, but no algorithm
has been widely adopted. (An investigation of this
problem is presented in [16].) One notable contact location choice algorithm, by Nguyen, uses an analytic
geometry approach to choose positions for opposing
grasps that are robust with respect to errors in position [31]. Some of the more successful autonomous
contact location choice systems are knowledge based
[23, 37].
4

Control Frameworks

The control of dexterous manipulation can be decomposed into three main levels, as shown in Table
2. High-level control includes task and motion planning and grasp choice. Most of these issues have been
discussed already in this paper. Mid-level control includes manipulation phases, for example whether the
ngers are operating independently or cooperatively,
and whether force or motion control is required. Transitions between these phases must be accomplished
smoothly, and events must be sensed to trigger the
transitions. Low-level control includes basic strategies,
(e.g. force or impedance control) and the formulations
of the control problem in the appropriate space (e.g.
the operational space of the grasped object).

Quality

Manipulators used for dexterous manipulation typically have kinematic redundancy. In addition, there
are usually multiple choices for contact locations that
achieve force closure on an object. Therefore, there
can be an innite number of possible grasps for a manipulation. We want to pick the \best" grasp. That
is, we want to choose the optimal contact locations,
contact forces, and nger poses for a particular manipulator, object and task combination.
In order to choose the best grasp, we need to develop a metric that will measure the \quality" of a
given grasp. It is common for this measure to depend on the task requirements. An example of such a
measure was developed by Li and Sastry [25], who separated the task requirements into two parts: wrench
(or force) requirement and twist (or motion) requirement. Each is represented by task ellipsoids, whose
axes indicate the relative magnitude requirements for
the elements of the wrench or twist vector.
While researchers have formulated good conceptual
quality measures for grasps, using these measures for
automatic grasp choice remains dicult. Many successful optimization techniques have been developed
for specifying contact forces given known contact locations and task requirements. Some of these are ecient enough for real time computation (for example,
[4]). Searching for the optimal contact locations is inherently more dicult, because the quality measure is

4.1

Mid-Level Control

The middle level of control for dexterous manipulation has received relatively little attention compared
to the high and low levels. One approach is proposed
by Hyde and Cutkosky [13]. The middle level of control involves the management of the various phases of
a task, as well as the specication of control laws to
be used when transitioning between phases. Research
in neurophysiology has shown that humans grasping
and manipulating objects use a similar approach, receiving signals from specialized skin cells that trigger
shifts between phases of a manipulation task [17].
Dierent phases are likely to use dierent control
laws, as dexterous manipulation is characterized by
changing kinematic and dynamic congurations and
5

Event A crosses
alert confidence

Contact Phase

Event A crosses
commitment confidence

Event B crosses
alert confidence

A
B

Commanded
Velocity

Event
Confidence

Free Motion Phase

Start-up
Action

Interaction
Force

Alert Action

Figure 7: Events and transitions between phases (adapted from [13]).

where (x) is the mass matrix, (x; x_ ) is the term
for centrifugal and Coriolis forces, p(x) is the term for
gravity, and F is the operational space force.
Using this framework, a control structure can be
chosen for dynamic decoupling and motion control.
For example, a basic PD (proportional-derivative) controller of the form

constraints. If the transitions between phases are not

managed carefully, these changes can cause undesirable behavior. Smoothness can be critical when event
detection relies on dynamic sensors, which are inherently sensitive to vibrations and small motion discontinuities. Smooth transitioning can be accomplished
by modifying the start and end portions of phases [13].
Figure 7 shows the how events are used to trigger transitions between phases, as well as the actions taken at
the beginning and end of a phase.
Knowing when a particular event, such as a nger
making contact with an object, has occurred is often
a dicult task in the presence of other events that
result in similar sensor signals. Thus, context-sensitive
event detection and sensor fusion are used to dene
the probability that a particular event has occurred
[9, 14, 38].
4.2

xd

kv (x_
kp
(xg
kv

_ )
x)
xd

(10)
(11)

may be used to move the operational point to a goal

position xg when a particular trajectory is not required.
The concept of impedance control is also important
in dexterous manipulation. In this type of control,
we specify the desired impedance of the object being
manipulated [10]. That is, the object in the grasp of
the robot can have an apparent mass, as well as stiness and damping when subjected to external forces.
The control block diagram in Figure 8 shows a control
framework using object impedance control for rolling
manipulation. This diagram shows the path of information from commands, through control laws, to application on the dexterous hand [14]. Another control
law, with explicit control of the contact trajectory, is
presented in [36].

Low-Level Control

Unlike the mid-level control strategies described

above, low-level control problems have received thorough attention in previous work. This is primarily
because the strategies for low-level control of dexterous manipulators are a direct extension of those used
for single and cooperating robot arms.
The formulation of the control has taken several
forms. One that stands out is the operational space
formulation, which has been used in object-centered
dexterous manipulation [20]. The basic idea of this
approach is to control motions and contact forces
through the use of control forces that act at the operational point of a grasped object. Through joint
space/operational space relationships formed using Jacobians and a Lagrange or Newton-Euler formulation,
an operational space dynamic model of the system is
created.
(x)x + (x; x_ ) + p(x) = F;

=
=

Manipulation Versus Exploration

As shown by Klatzky and Lederman [21], manipulation and exploration go hand in hand. We can obtain
a precise denition for each separately: \Pure" manipulation occurs when the object is completely known.
\Pure" exploration happens when the object is xtured and is not known. Most dexterous manipulation is a combination of manipulation and exploration.

(9)
6

Commands

Control Laws

Desired
Body
Position

Object
Impedance
Controller

Desired
Contact
Locations

Rolling
Controller

Desired
Internal
Force

Internal
Force
Controller

Application and Feedback

Contact
Point
Sensing

Backward
Grasp
Transform

Forward
Grasp
Transform

Finger
Control

Plant

Figure 8: Control block diagram for manipulation with rolling.

And in a non-ideal world, we need manipulation for
exploration and vice versa.
There are a number of examples of simultaneous
manipulation and exploration. Bicchi, et al. [3] manipulate by rolling and also build up a model of the
object at the same time. Okamura, et al. manipulate and explore an object where the general shape is
assumed but small features may be explored [34, 33].
Allen primarily uses exploration with robot ngers to
build up an object model using mathematical shape
descriptions (superquadrics) [1].

manipulation/haptic exploration [2]). Tactile sensors

can be divided into two categories: extrinsic and intrinsic. Extrinsic sensors can sense actual contact location. They are often array sensors and are limited in
resolution and accuracy. Examples include capacitive
arrays, pin arrays, micromachined force sensor arrays,
and optical measurement systems (i.e., CCD). Problems with these sensors include sensitivity to noise,
delicacy, poor resolution, slow data acquisition and
processing, diculties in manufacturing, and high cost
[11].
Intrinsic sensors use global measurements from
which contact information can be extracted. Types
of intrinsic sensors include force/strain sensors, optical sensors (i.e. PSD), and joint torque sensing. While
some progress has been made, these sensors are often
too large for dexterous robotic ngers, are expensive
for multiple degrees of freedom, and provide limited information. For example, the optical waveguide tactile
sensor designed by Maekawa, et al. gives only the centroid of all contact locations, which is aected by the
intensity of the dierent contacts [26]. Intrinsic sensors also cannot tell the dierence between one contact
and multiple contacts.

6 Current Limiting Factors

Several robotic hands have been developed in the

past two decades, although at the time of this writing
they are used mainly in research laboratories. Limitations in hardware and in software (algorithms) hamper
their application outside the laboratory.
6.1

Hardware

High bandwidth and well-controlled actuation is

necessary for dexterous manipulation. In addition, actuators should be small and lightweight in comparison to the hand they are actuating. Currently, small
but powerful actuators are rare and expensive. However, new technologies such as voice-coil actuation and
rare earth magnets are enabling smaller construction.
Backdrivability is another issue; in order to increase
power, designers of robotic hands often use transmission systems that cause small motion errors to result
in high interaction forces, reducing backdrivability.
Sensors are essential for dexterous manipulation, as
feedback of contact information is vital to accomplish
any task where the environment in unknown. Tactile
sensors are the primary sources of information for most
dexterous manipulation tasks (although some work has
been done on combining vision sensors with dexterous

6.2

Software

The mathematical complexity involved in spatial

rolling and sliding manipulations has discouraged the
use of 3-D multi-ngered hands. While some investigators do work with 3-D manipulators [36], others have
chosen to limit dexterous manipulation to 2-D. Another approach is to design a simplied manipulator
that uses planar ngertips for 3-D rolling [3, 8].
There is certainly a gap between theoretical promise
and practical delivery due to the complexity of specifying and controlling automated grasping and manipulation tasks. There are several areas of dexterous manipulation in which better algorithms are required before
signicant improvement can be made. Currently, most
7

grasp choice and optimization systems that use multiple ngers are quite slow and the calculations must
be done o-line, particularly when contact locations
must be determined. Another area for improvement
is motion planning. Similar to grasp choice, the algorithms are slow and cannot be accomplished during
the manipulation task. One nal area is the use of
tactile sensing in control. Understanding and using
tactile sensor output for direct servoing has been the
subject of some recent work [33, 27, 43], however, improvements in tactile sensing and data interpretation
are needed to accomplish this in less controlled conditions.

[3] Antonio Bicchi, Alessia Marigo, and Domenico

Prattichizzo, Dexterity through rolling: Manipulation of unknown objects, Proceedings of the
IEEE International Conference on Robotics and
Automation (1999), 1583{1588.
[4] Martin Buss, Hideki Hashimoto, and John B.
Moore, Dextrous hand grasping force optimization, IEEE Transactions on Robotics and Automation 12 (1996), no. 3, 406{418.
[5] Mark R. Cutkosky, Robotic grasping and ne manipulation, Kluwer Academic Publishers, 1985.
[6] Mark R. Cutkosky and Robert D. Howe, Human
grasp choice and robotic grasp analysis, Dextrous
Robot Hands (Subramanian T. Venkataraman
and Thea Iberall, eds.), Springer-Verlag, 1990,
pp. 5{31.

7 Discussion

At present, autonomous, real time dexterous manipulation in unknown environments still eludes us.
In much of the current research, it appears that we
have given up on anthropomorphic hands because of
diculties in hardware development and autonomous
control. The recent trend has been to break the dexterous manipulation problem into small parts that can
be studied separately with specialized hardware. In
many cases the research is done in simulation rather
than experimentally.
Although autonomous dexterous manipulation remains impractical outside of the laboratory,a promising interim solution is supervised manipulation. In
this approach, a human provides the high-level grasp
and manipulation planning, while the robot performs
ne (dexterous) manipulations [32]. Another method
is teaching by demonstration (gesture-based programming) [40]. The human may also perform the interpretation of tactile information in supervised remote
exploration [32].
The miniaturization of manipulation is another area
of with promise. However, manipulations occurring on
a very small scale are dominated by friction and Van
der Waals forces. Stable grasping is often not necessary; the objects will stick directly to the manipulator.

[7] Mark R. Cutkosky and Paul Wright, Friction, stability and the design of robot ngers, International
Journal of Robotics Research 5 (1986), no. 4, 20{
37.
[8] P. Datseris and W. Palm, Principles on the dvel-

opment of mechanical hands which can manipulate objects by means of active control, Journal of

Mechanisms, Transmissions, and Automation in

Design 107 (1985), no. 2, 148{156.

[9] B. Eberman and J. Kenneth Salisbury, Applica-

tion of change detection to dynamic contact sensing, International Journal of Robotics Research

13

(1994), no. 5, 369{394.

[10] Neville Hogan, Impedance control: an approach to

manipulation: parts i, ii, and iii, ASME Journal
of Dynamic Systems, Measurement, and Control
107 (1985), 1{24.
[11] Robert D. Howe and Mark R. Cutkosky, Touch
sensing for robotic manipulation and recognition,
The Rbotics Review 2 (O. Khatib, J. Craig, and
T. Lozano-Perez, eds.), MIT Press, 1992, pp. 55{
112.
[12]

References

, Practical force-motion models for sliding

mnaipulation, International Journal of Robotics
Research 15 (1996), no. 6, 557{572.

[1] Peter K. Allen and Paul Michelman, Acquisition

and interpretationof 3-d sensor data from touch,
IEEE Transactions on Robotics and Automation
6 (1990), no. 4, 397{404.

[13] James M. Hyde and Mark R. Cutkosky, Phase

[2] Peter K. Allen, Andrew T. Miller, Paul Y. Oh,

and Brian S. Leibowitz, Using tactile and visual
sensing with a robotic hand, Proceedings of the
IEEE International Conference on Robotics and
Automation (1997), 676{681.

[14] James M. Hyde, Marc Tremblay, and Mark R.

Cutkosky, An object-oriented framework for
event-driven dexterous manipulation, Experimental Robotics IV, Springer-Verlag, 1995.

management framework for event-driven dextrous

manipulation, IEEE Transactions on Robotics

and Automation 14 (1998), no. 6, 978{985.

[15] Thea Iberall, A neural network for planning

hand shapes in human prehension, Proceedings of
the 1988 Automatic Controls Conference (1988),
2288{2293.
[16] John W. Jameson, Analytic techniques for automated grasp, Ph.D. thesis, Stanford University,
Department of Mechanical Engineering, 1985.
[17] R.S. Johansson and G. Westling, Aerent signals during manipulative tasks in man, Information Processing in the Somatosensory System:
Proceedings of an International Seminar at the
Wenner-Gran Center (O. Franzen and J. Westman, eds.), Macmillan, 1991.
[18] Imin Kao and Mark R. Cutkosky, Quasistatic manipulation with compliance and sliding, International Journal of Robotics Research 11 (1992),
no. 1, 20{40.
[19] Jerey Kerr and Bernard Roth, Analysis of multingered hands, International Journal of Robotics
Research 4 (1986), no. 4, 3{17.
[20] Oussama Khatib, Unied approach for motion

[27]

tional Conference on Robotics and Automation

(1995), 743{750.
[28] Matthew T. Mason and J. Kenneth Salisbury,
Robot hands and the mechanics of manipulation,
The MIT Press, 1985.
[29] D.J. Montana, The kinematics of contact and
grasp, International Journal of Robotics Research
7 (1988), no. 3, 17{32.
[30] R. Murray, Zexiang Li, and Shankar Sastry, A
mathematical introduction to robotic manipulation, CRC Press, 1994.

[31] Van-Duc Nguyen, Constructing force closure

grasps, International Journal of Robotics Research 7 (1988), no. 3, 3{16.
[32] Allison M. Okamura, Michael A. Costa,
Michael L. Turner, Christopher Richard,
and Mark R. Cutkosky, Haptic exploration of
surfaces, Experimental Robotics VI (Peter Corke
and James Trevelyan, eds.), Lecture Notes in
Control and Information Sciences, vol. 250,
Springer-Verlag, 2000, ISBN: 1 85233 210 7,
pp. ??{??
[33] Allison M. Okamura and Mark R. Cutkosky, Haptic exploration of ne surface features, Proceedings of the IEEE International Conference on
Robotics and Automation 3 (1999), 2930{2936.
[34] Allison M. Okamura, Michael L. Turner, and
Mark R. Cutkosky, Haptic exploration of objects
with rolling and sliding, Proceedings of the IEEE
International Conference on Robotics and Automation 3 (1997), 2485{2490.
[35] Daniela Rus, Analysis of musical-instrument
tones, International Journal of Robotics Research
18 (1999), no. 4, ??{??
[36] N. Sarkar, X. Yun, and V. Kumar, Dynamic con-

and force control of robot manipulators: the operational space formulation, IEEE Journal of

[21]

[22]

[23]

[24]

[25]

[26]

, Tactile sensor based manipulation of

an unknown object by a multingered hand with
rolling contact, Proceedings of the IEEE Interna-

Robotics and Automation 3 (1987), no. 7, 43{53.

Robert L. Klatzky and Susan Lederman, Intelligent exploration by the human hand, Dextrous
Robot Hands (Subramanian T. Venkataraman
and Thea Iberall, eds.), Springer-Verlag, 1990,
pp. 66{81.
Vijay Kumar and Antonio Bicchi, Review paper for contact and grasp symposium, Symposium
Proceedings of the IEEE International Conference
on Robotics and Automation (2000, in review).
C. Laugier and J. Pertin, Automatic grasping: A case study in accessibility analysis, Advanced Software in Robotics (A. Danthine and
M. Geradin, eds.), Elsevier Science Publishers,
1984, pp. 201{214.
Susanna R. Leveroni, Grasp gaits for planar object manipulation, Ph.D. thesis, Massachusetts
Institute of Technology, Department of Mechanical Engineering, 1997.
Zexiang Li and Shankar S. Sastry, Issues in dextrous robot hands, Dextrous Robot Hands (Subramanian T. Venkataraman and Thea Iberall, eds.),
Springer-Verlag, 1990, pp. 154{186.
H. Maekawa, K. Tanie, and K. Komoriya,

trol of 3-d rolling contacts in two-arm manipulation, IEEE Transactions on Robotics and Au-

tomation 13 (1997), no. 3, 364{376.

[37] Sharon Stanseld, A robotic perceptual system
utilizing passive vision and active touch, International Journal of Robotics Research 7 (1988),
no. 6, 138{161.
[38] Marc R. Tremblay, Using multiple sensors and

A nger-shaped tactile sensor using an optical

waveguide, Proceedings of the IEEE International

contextual information to detect events during a

manipulation task, Ph.D. thesis, Stanford Uni-

Conference on Systems, Man and Cybernetics

(1993), 403{408.

versity, Department of Mechanical Engineering,

1995.
9

[39] Francisco Valero-Cuevas, Clinical applications of

dexterous manipulation and grasp, Symposium
Proceedings of the IEEE International Conference
on Robotics and Automation (2000, in review).
[40] Richard M. Voyles and Pradeep K. Khosla,
Gesture-based programming:
a preliminary
demonstration, Proceedings of the IEEE Interna-

tional Conference on Robotics and Automation

(1999), 708{713.
[41] David Williams and Oussama Khatib, The virtual

linkage: a model for internal forces in multi-grasp

manipulation, Proceedings of the IEEE Interna-

tional Conference on Robotics and Automation 1

(1993), 1025{1030.
[42] Tsuneo Yoshikawa and Kiyoshi Nagai, Analysis
of multi-ngered grasping and manipulation, Dextrous Robot Hands (Subramanian T. Venkataraman and Thea Iberall, eds.), Springer-Verlag,
1990, pp. 5{31.
[43] H. Zhang, H. Maekawa, and K. Tanie, Sensitivity analysis and experiments of curvature estimation based on rolling contact, Proceedings of the

IEEE International Conference on Robotics and

Automation (1996), 3514{3519.

10